The Political Economy of Bolivia s New Regionalism: A Look at Electoral Patterns in Santa Cruz, Tarija, and Chuquisaca

Transcription

1 University of Mississippi From the SelectedWorks of Miguel Centellas May, 2009 The Political Economy of Bolivia s New Regionalism: A Look at Electoral Patterns in Santa Cruz, Tarija, and Chuquisaca Miguel Centellas Miguel A. Buitrago Available at:

2 The Political Economy of Bolivia s New Regionalism: A Look at Electoral Patterns in Santa Cruz, Tarija, and Chuquisaca Miguel Centellas Assistant Professor of Political Science Mount St. Mary s University Old Emmitsburg Road Emmitsburg, MD centellas@msmary.edu Miguel A. Buitrago Research Associate/Doctoral Candidate GIGA Institut für Lateinamerika-Studien Neuer Jungfernstieg Hamburg, Germany buitrago@giga-hamburg.de Abstract While most conventional accounts assume an economic motivation behind the new regional autonomy movements, such claims have not been systematically tested. In this paper, we test the link between electoral support for Evo Morales and regional autonomy to socioeconomic and demographic indicators at the municipal level. Our cases include the three gas-producing departments of Santa Cruz, Tarija, and Chuquisaca. Our unit of observation is the municipality, for a total of 95 units. We look at eight elections, taking place at five distinct moments, between December 2005 and January We test whether voting districts with higher levels of socioeconomic indicators or lower percentages of indigenous populations were more likely to vote for autonomy, for the prefect, and against Evo Morales and MAS. Our findings raise questions about the validity of conventional claims that regional autonomy movements in the so-called Media Luna are primarily driven by socioeconomic inequalities or ethnic cleavages. Prepared for delivery at the Annual Meeting of the Midwest Political Science Association, Chicago, April 2-5, 2009.

3 The Ongoing Debate over Regional Autonomy On January 28, 2005, Santa Cruz celebrated, what was going to become, a historic town hall meeting. The prefect of the Santa Cruz department, now self proclaimed governor, Rubén Costas, asked the hundreds of thousands of citizens gathered at the Cristo Redentor square if they agreed to begin with the process that was going to lead to more autonomy (i.e. self-government and self-determination). The citizens accepted with a resound yes and with that the autonomic movement of Santa Cruz began its work. In the context of a deeply polarized society the issue of autonomy has become a major point of discordance and the source of an ongoing polemic debate. While Santa Cruz, primarily, has been leading the efforts for a more decentralized government and now for autonomy based on the inefficiencies of a centralized government and its effects on regional development, there are those who counter those contentions based on ongoing power struggle, elite interests and social structures arguments. A coincidence in this debate, however, has been the criticism of the modern Bolivian nation state. This debate, which has been carried out very publically not only in the political arena, but also in the Bolivian media is reflected in the scholarly work among Bolivian researchers. This brief discussion of the current debate on autonomy does not aim at an exhaustive review of that debate, but merely seeks to reflect its central ongoing aspects. At this point in time, the debate can be divided into three different approaches. One approach takes into account the political and institutional aspects; a second approach looks at the economic reasons; and a third approach seeks to explain the developments leading to the autonomic demands. All of these approaches however touch on issues of identity, exclusion, and self-determination. At the heart of the political approach is the questioning of the nature of the Bolivian state. Maria Zegada (2007) takes this approach and argues that the two dominant and often overlapping clusters framing the debate are the civic-regional and indigenous-peasant discourses. These highlight the deep socioeconomic, politico-institutional, ethno-cultural, and territorial division existing in Bolivian society. Other scholars take a historic look and go back to colonial times to argue that the current form of the state is based on colonial designs brought by the Spanish crown. This system was based on segregationist structures and was designed to render all benefits to the conquerors. Throughout the new republic, the same system was continued by the local elites to perpetuate the same benefits. According to this line of argumentation, to this day, the calls for autonomy from the departments is just the effort to preserve the status quo in favor of the elites (Garcia 2007; Tapia 2007). At the same time, this approach argues for an indigenous autonomy, which is supposed to lead to the decolonization of the state (Garcia 2007). Namely, the dismantling of the old segregationist structure to build an inclusive system where the indigenous peoples can participate in the decision making process with their own traditional organizational and power structures (Garcia 2007, Tapia 2007). The indigenous peoples have a right to selfgovernment and self-determination. The departments of Santa Cruz, Tarija, Beni, and Pando further their argument under the same lines of political argumentation: self-government and self-determination. For scholars advancing this side of the debate, the same centralist governmental system inherited from colonial and post-colonial times has been the problem acting against regional development. Roca (2007) takes a historical approach, reminding us that the struggle against centralism begins already in the 19th century. A vivid reminder is the 1

4 1898 Federalist War, when the seat of government was moved from Sucre to La Paz under the disguise of federalist ideals. Furthermore, he reminds us that the decentralization efforts in Santa Cruz begin in early 20th century, with the campaign gaining an eleven per cent taxes for Santa Cruz out of oil exports, and continue to this day. Roca contends that this struggle for decentralization and now for autonomy is elite driven, and that in the background are economic incentives. Autonomy should help the regions address their problems by themselves without having to wait for La Paz s agreement (Roca 2007). Parallel to the latter are issues such as the advantages and disadvantages of autonomy, regional development. The debate addressing this question has been largely carried out during the constitutional assembly process in 2006/2007. A diverse number of political parties, interest groups, and individuals argued for and against autonomy. This debate reflected the political orientations of the government and its supporters, on the one side, and of the departmental governments (primarily Santa Cruz, Tarija, Beni, and Pando), the civic organizations, and its supporters, on the other side. Although similar in nature, the two positions had fundamental differences. The government proclaimed a kind of autonomy based on ethno-cultural and identity factors, such as indigenous communities or indigenous tribes. The regional autonomic proposal emphasized the crucial role of the departmental government as, not only the pillar for development, but the source of self-determination (Zegada 2007). A second approach highlights the economic motivations to achieve autonomy. Although, this approach uses the same lines of argumentation as the first approach, it concentrates on the economic incentives. There are those who argue that regional governments take the autonomic issue as the means to gain total control over the decision making process, and thus over the financial gains, in the regional governments (Roca 2007). Furthermore, the elites are the one force seeking the control to keep the status quo (Mansilla 2007). In the same manner, Isaac Sandoval (1985) and more recently Blanes (2007) argue that the autonomic push in Santa Cruz is motivated by economic and political reasons, such as the rents from oil and now natural gas production and more decision power. They contend that the principal actors were and still are the civic committees in each department, which have historically represented regional interests. A third approach to the debate is the attempt to explain the autonomic movement. In that manner we find researchers such as Seleme (2007) who takes a constructivist approach arguing that the elites in Santa Cruz have been constructed by different factors, such as geographic (oriente), historic (forgotten region), culture (camba), and ideological (cruceño). This interpretation should lead us to understand why the elites in Santa Cruz seek autonomy from the central government in La Paz. Albó and Barrios (2007) discuss the multi and pluri nature of the indigenous groups in the Andes. Their approach contributes to the understanding of why the indigenous peoples seek autonomy. In the end, they propose three types of autonomies: departmental, indigenous, and municipal. One year after the Constitutional Assembly ended its work and two months after the new constitution was accepted in the January 25, 2009, referendum, the debate over autonomy in Bolivia has not ended. In fact, in light of the current political situation, it has intensified due to the contrasting points of view among people who support a departmental level of autonomy and those who appeal at a localized ethnic level of autonomy. This issue is not yet resolved and will be the source of more political strife 2

5 among the two antagonist groups. However, the current debate has taken a turn towards reconsidering the reasons for autonomy and whether the constitutional assembly process is valid or not (for this debate see Reyes 2008; Molina 2008; Hurtado 2008; PIEB 2008; Garcia 2007; Leaño 2007; Tapia 2007). Our paper seeks to contribute to this debate by empirically testing some of the implicit assumptions about the electoral support for Evo Morales and MAS (and their indigenous-popular reform movement), on the one hand, and the opposition prefects and regional autonomy movements on the other. Research Design Our study looks at only three of Bolivia s nine departments: Santa Cruz, Tarija, and Chuquisaca. A focus on these three cases is appropriate because we are primarily interested in studying support for regional autonomy in Media Luna departments, defined as departments that have recently developed regional autonomy movements that challenge the government of Evo Morales and MAS. Both Santa Cruz and Tarija have powerful, well-developed, and growing regional autonomy movements dating back at least to the December 2005 prefect elections, when voters there elected opposition prefects (Ruben Costas in Santa Cruz and Mario Cossío in Tarija). We include Chuquisaca in our study because it has more recently developed a new and peculiar regional autonomy movement. Like in Santa Cruz and Tarija, the movement s base has come from the departmental capital. But unlike in Santa Cruz and Tarija, Sucre s drive for regional autonomy grew from a particular grievance based on the city s demand to move the political capital back to the country s historic (and constitutional ) capital city. The capitalía issue introduced a new challenge to Morales's government; it also seems to have split the department along an urbanperiphery cleavage. A special election for prefect in June 2008 saw the election of an indigenous female prefect (Savina Cuéllar) who had recently split with MAS and was endorsed by Sucre s elite establishment. Case Selection Our study seeks to look at three important, yet culturally different, departments that represent a particular sub-section of Bolivia s political geography. Each department also includes a section of the Bolivian Chaco, the oil and natural gas rich region of the country that is often viewed as lying at the heart of the autonomy conflict. Our hope is to understand what factors can help explain support for regional autonomy in Bolivia's eastern departments. Santa Cruz. According to the most recent census (2001), the department of Santa Cruz has at least 2.03 million inhabitants, of which approximately three quarters live in urban areas; this includes the 1.13 million who live in the city of Santa Cruz, the department capital. Historically, the department of Santa Cruz has been the region with the longest aspiration for autonomy and a more decentralized state. Interestingly, the chief protagonists for such demands have rarely been political parties, but rather the region s developed civic organizations (principally, the Comité Cívico pro Santa Cruz). Based in the city of Santa Cruz, the Comité Cívico has been a principle historical force in the region s cultural self-articulation (see P. Peña et al 2003). 3

6 In the wake of the 1952 National Revolution and the state-led development of oil and gas fields in the region, civic organizations sought to defend the department s regional interests. A major political gain and a product of an active social movement was the passing of law guaranteeing that eleven percent of state oil and gas revenues would remain in the department that produced them. Ultimately, the national revolution left a positive balance for Santa Cruz, which has experienced exponential growth in the subsequent decades. It also set the basis for agro-industrial activity and better integrated (a stated goal of the MNR-led revolution) the previously peripheral region into the political and economic life of nation-state. By 2004, Santa Cruz accounted for nearly a third of the country s total GDP, making it the single largest contributor to the country s economy, including nearly a quarter of all petroleum and natural gas production (INE 2004). Tarija. According to the most recent census, the department of Tarija has more than 390,000 inhabitants, of which nearly two thirds live in urban areas; this includes most of the more than 150,000 who live in the city of Tarija, the department capital. Like Santa Cruz, Tarija has a long history of defining itself as culturally different from Andean Bolivia, though in a more ambivalent way (see L. Peña et al 2003). Unlike Santa Cruz, however, Tarija did not develop a significant autonomy political-social movement until very recently; in doing so, however, it has closely aligned itself with the Santa Cruz movement (which is also influential in movements in Beni and Pando departments). Despite its small size, Tarija is significant economically. With only slightly more than four percent of the country s population, it contributes more than eight percent of the country s GDP. Additionally, most of the country s natural gas reserves are located in the department s eastern provinces (Gran Chaco and O Connor). In 2004, Tarija alone produced more than half of the country s share of petroleum and natural gas production (INE 2004). Chuquisaca. According to the most recent census, the department of Chuquisaca has more than 530,000 inhabitants, of which only about 40 percent live in urban areas; this includes most of the more than 100,000 who live in the city of Sucre, the department capital. Unlike Santa Cruz or Tarija, Chuquisaca has historically identified itself as an Andean part of Bolivia, much like La Paz. In contrast, the department has long identified itself as an integral part of the Bolivian nation, harkening back to Sucre s role as the country s first capital and the birthplace of national (and also Spanish American) independence. Chuquisaca or, more accurately, Sucre has only recently developed a regional autonomy movement closely aligned to those in Santa Cruz and Tarija. This movement, however, emerged only during the 2006 constituent assembly process, as a (primarily) Sucre-based movement demanded that the constituent assembly discuss a proposal to move the country s capital from La Paz back to Sucre. A series of violent clashes between capitalía supporters and pro-mas groups led the MAS prefect to resign; a special by election saw Savina Cuéllar, a Quechua-speaking, indigenous, female pro-capitalía candidate. win the prefecture in an election that deeply divided the department along city and rural cleavages. Despite its historical importance, Chuquisaca is in most areas an economic lightweight. Its share of the country s GDP is less than five percent (about equal to its 4

7 share of the population). Even though two of its provinces (Luis Calvo and Hernando Siles) are located in the oil and gas producing Chaco regions, Chuquisaca produced less than four percent of the country s oil and natural gas, though oil and gas production is expected to expand in the region. Hypotheses Our hypotheses aim to test the conventional wisdom, which assumes that regional movements in the Media Luna are driven by socioeconomic and ethnic cleavages, which overlap and re-enforce each other. The conventional interpretation of contemporary Bolivian politics is that poor, rural, and indigenous voters are more likely to support Morales and MAS, are less likely to support regional autonomy, and are more likely to support the new constitution approved by voters in January The following operational hypotheses predict support across municipalities for MAS, defined as support for Morales, MAS party candidates, and the new draft constitution: Hypothesis 1a: As the percent of rural population in a municipality increases, its support for MAS increases. Hypothesis 1b: As the rate of poverty in a municipality increases, its support for MAS increases. Hypothesis 1c: As the percent of self-identified indigenous population in a municipality increases, its support for MAS. Hypothesis 1d: As the rate of literacy in a municipality increases, its support for MAS decreases. The following operational hypotheses predict support across municipalities for regional autonomy, defined as support for pro-autonomy prefect candidates and votes for autonomy in regional referenda: Hypothesis 2a: As the percent of rural population in a municipality decreases, its support for regional autonomy increases. Hypothesis 2b: As the rate of poverty in a municipality decreases, its support for regional autonomy increases. Hypothesis 2c: As the percent of self-identified indigenous population in a municipality decreases, its support for regional autonomy increases. Hypothesis 1d: As the rate of literacy in a municipality increases, its support for regional autonomy increases. Data and Method Our paper uses municipal-level data from two Media Luna departments (Santa Cruz and Tarija), as well as one department that has only recently developed an active regionalist movement (Chuquisaca). The data is a mix of electoral data covering the period from the December 2005 general and prefect elections and the January 2009 constitutional referendum. In addition to electoral data, we also use socioeconomic and demographic indicators from the 2001 census (the most recent census data available). This gives us 95 possible observations (municipalities) across all three departments. 5

8 Socioeconomic and Demographic Variables. We rely on three basic socioeconomic variables: ruralness, poverty, and literacy. Although the three are often associated (poverty, ruralness, and illiteracy are associated), we find substantial variation across the municipalities in our three departments. Each of these variables is a ratio measure, ranging from zero to one. The degree of ruralness in each municipality is the percentage of municipal residents who are categorized as rural by Bolivia s Instituto Nacional de Estadísticas (INE). The level of poverty in each municipality is the percent of municipal residents whose basic needs unsatisfied, as defined by the Bolivian government s Unidad de Análisis de Políticas Sociales y Económicas (UDAPE). The literacy rate in each municipality is calculated from the illiteracy rate (as one minus the illiteracy rate) reported by UDAPE. Unlike the absolute literacy rate reported by INE, the UDAPE illiteracy rate specifically looks at the population aged 15 to 44. In addition to simple socioeconomic factors that might affect voting in our three departments, we are also interested in the effect indigeneity may have in explaining electoral differences across municipalities. Our variable for municipal indigeneity is the percentage of the population 15 years or older who self-identify as belonging to one of the various indigenous communities in the 2001 census. Because the 2001 census took place before six Santa Cruz municipalities (San Juan de Yapacaní, La Bélgica, Puerto Fernández, San Pedro, Cuatro Cañadas, and El Carmen) were created, we do not have figures for these municipalities. This drops our number of observations (N) for models using indigeneity as a variable from 95 to 89. Data for our socioeconomic figures are available, however, from UDAPE and/or INE, using 2001 census data. One of the remarkable things to note about the municipalities in our three departments is their broad range across our socioeconomic and demographic variables (see Table 1). The broadest range is actually in Santa Cruz, which has eleven municipalities that are 100 percent rural and three that are less than 5 percent rural. Perhaps surprisingly, Santa Cruz has a broad range of indigeneity as well: The capital city of Santa Cruz is 32.1 percent indigenous, while the range across the department is from 92.7 percent (Pailón, about an hour east of the capital) to 1.6 percent (Samaipata, about an hour west of the capital). There is also little overlap between our socioeconomic and demographic variables: The municipality of Postrervalle is 100 percent rural, 4 percent indigenous, and 85 percent poor. The municipality of Quijaro is only 3 percent rural, but 51 percent indigenous and 24 percent poor. The municipality of Okinawa is 65 percent rural, 34 percent indigenous, and 54 percent poor. The same holds true even in Chuquisaca s 28 municipalities. Surprisingly, the capital (Sucre) has a higher share of indigenous population than twelve municipalities; the Chuquisaca municipality with the smallest share of indigenous population is Las Carreras (which borders Potosí), which is nevertheless 100 percent rural and 89 percent poor. 6

9 Table 1. Socioeconomic indicators by municipalities (as percentage) Chuquisaca Tarija Santa Cruz Rural population Highest Lowest Department capital Indigenous population Highest Lowest Department capital Poor population Highest Lowest Department capital Literate population Highest Lowest Department capital Source: INE and UDAPE. Elections and Political Variables. Our study covers eight different elections, which took place at five distinct moments in time: December 2005 general election (Santa Cruz, Tarija, Chuquisaca) December 2005 prefect election (Santa Cruz, Tarija, Chuquisaca) on the same day as the general election but on a separate ballot July 2006 constituent assembly election (Santa Cruz, Tarija, Chuquisaca) July 2006 autonomy referendum (Santa Cruz, Tarija, Chuquisaca) on the same day as the constituent assembly election but on a separate ballot May/June 2008 autonomy referendums (Santa Cruz and Tarija) June 2008 Chuquisaca prefect election August 2008 recall referendums (Santa Cruz, Tarija, and Chuquisaca 1 ) January 2009 constitutional referendum We were able to obtain electoral data disaggregated to the municipal level for each of the eight elections under study. Data for each of these elections come from Bolivia's Corte Nacional Electoral (CNE), with the exception of the 2008 autonomy referendums. Because the May 2008 (Santa Cruz) and June 2008 (Tarija) autonomy referendums were not sanctioned by the central government, these wildcat referendums were organized entirely by their respective department's regional electoral court. Data for each of these 1 Chuquisaca voters only voted in the presidential referendum, not the prefect referendum. They had only months earlier elected a prefect in a special by election. 7

10 two referendums come from the website of the respective department s Corte Departamental Electoral (CDE). We looked at a number of specific political-electoral variables, each meant to assess one of our two primary dependent variables: support for Morales and/or MAS and support for regional autonomy. For each of the elections, we looked at the following set of operational dependent variables: In the December 2005 general election, we specifically looked at votes for Morales, who headed the plurinominal list in each department. We did not look at municipal-level uninominal votes (votes for party candidates in single-member districts). We chose instead to look at a normalized vote choice that had a consistent slate of candidates across the entire department. In the December 2005 prefect election, we specifically looked at votes for MAS candidates. We chose to do this, rather than look at votes for opposition candidates for a number of reasons. First, opposition candidates only won in Santa Cruz and Tarija. Second, in each of those departments MAS candidates placed third in contests where both opposition (non-mas) challengers favored regional autonomy. In the July 2006 constituent assembly election, we specifically looked at votes for the MAS departmental electoral list. As with the 2005 general election, we chose to look only at a slate of candidates for which all voters in a department would vote. We thus ignored lists specific to uninominal districts. In the July 2006 autonomy referendum, we merely looked at the share of votes favoring autonomy. In the May/June 2008 autonomy referendums, we also merely looked at the share of votes favoring autonomy. In the June 2008 Chuquisaca prefect election, were merely looked at the share of votes for Savina Cuéllar. In the August 2008 recall referendums, we looked at two variables. The first was votes for Morales. The second was votes for incumbent prefect. The latter did not apply to Chuquisaca, since voters there did not cast prefect recall ballots. In the January 2009 constitutional referendum, we merely looked at the share of votes in favor of the proposed constitution. In addition, a number of these political-electoral variables were later used as independent variables. This allowed us to control for any effects from previous elections. For example: We knew that municipalities that had voted for autonomy in 2006 would again support autonomy in But because municipalities that had voted against autonomy in 2006 voted for autonomy in 2008, and because vote shares had shifted, we wanted to determine whether socioeconomic or demographic factors had played any role in that shift. We also wanted to know whether there was any relationship between support for autonomy and support for Morales or opposition prefects. Including results from other vote choices in models allowed us to test for such relationships. Finally, we also looked at models of voter turnout in a number of key elections, to determine whether socioeconomic, demographic, or political-electoral factors played any role in determining voter turnout. This allowed us to statistically test whether (consistent with conventional wisdom) voter turnout was lower in more rural, poorer, less educated, and more indigenous municipalities. 8

11 Analysis To test our various hypotheses, we developed a number of simple multivariate regression models in which all our variables were ratio measures (ranging from 0 to 1). Overall, our results were surprising and counter-intuitive. Socioeconomic and demographic variables rarely mattered in most of our models. Instead, we found that political-electoral models were more significant and often wiped out any effect from our independent variables. The 2005 General and Prefect Elections Our first set of models looked at voter turnout across all three departments (see Table 2). Neither ruralness nor poverty could explain voter turnout differences in the 2005 general election in any of our models. Literacy increased voter turnout slightly, but only in models that did not include indigeneity. Our final model did see a statistically significant effect of indigeneity on voter turnout, but the effect was very small (a one percent increase in the number of indigenous population produced a minuscule drop in voter turnout). Additionally, none of the models were very strong, with low R-square values. Table 2. Regression estimates of voter turnout in the 2005 general election at the municipal level in Santa Cruz, Tarija, and Chuquisaca departments Model A Model B Model C Rural Poor Literacy * Indigenous * Constant ** ** ** R-squared N observations When attempting to estimate support for Morales in the 2005 general election across all three departments (see Table 3), we found no evidence that any of our independent variables had any significant effect on vote differences across municipalities. Only in department-specific models (see Table 4) did we find any support for our research hypotheses, but only in Chuquisaca. There, literacy did significantly decrease support for Morales, even as indigeneity increased his support. Surprisingly, the Chuquisaca model has a relatively high R-square value despite an N of 28. Surprisingly, levels of poverty, ruralness, literacy, or percent indigenous population had no statistically significant effect in explaining the level of support for Morales in Santa Cruz or Tarija. Even though Morales lost the department vote in both Santa Cruz and Tarija (he won in Chuquisaca), we would have expected that rural, poor, and indigenous voters would have supported him; there is no evidence of this in our models. 9

12 Table 3. Regression estimates of support for Evo Morales in the 2005 general election at the municipal level in Santa Cruz, Tarija, and Chuquisaca departments Model A Model B Model C Rural Poor Literacy Indigenous Constant ** ** ** R-squared N observations Table 4. Regression estimates of support for Evo Morales in the 2005 general election at the municipal level in department-specific models Chuquisaca Tarija Santa Cruz Rural Poor Literacy ** Indigenous ** Constant ** R-squared N observations Turning to support for MAS prefect candidates in 2005, we find stronger support for the effect of at least some of our independent variables (see Table 5). When looking at all three departments, we find a strong and statistically significant effect for literacy. For every one percent increase in the literacy rate, our model estimates an almost similar decrease in support for MAS prefect candidates. Because the MAS candidate was successful only in Chuquisaca (where David Sánchez won with 42.3 percent of the department-wide vote), we ran separate models looking at only Santa Cruz and Tarija (where the MAS candidate lost) and looking only at Chuquisaca. In Santa Cruz and Tarija, only literacy mattered, but it mattered a great deal: a one percent increase in literacy was correlated to an estimated two percent decrease in support for the MAS candidate. In the Chuquisaca-only model, both indigeneity and literacy mattered, in roughly the same proportions as they had when looking at support for Morales. 10

13 Table 5. Regression estimates of support for MAS candidate in the 2005 prefect election at the municipal level Santa Cruz, Tarija, and Chuquisaca Santa Cruz and Tarija Chuquisaca Rural Poor Literacy ** * ** Indigenous ** Constant ** ** ** R-squared N observations Table 6. Regression estimates of support for MAS candidate in the 2005 prefect election at the municipal level (including support for Morales on presidential ballot) Santa Cruz, Tarija, and Chuquisaca Santa Cruz and Tarija Chuquisaca Rural * Poor Literacy ** * Indigenous Support for Evo on * ** presidential ballot Constant ** ** R-squared N observations When we looked at similar models that included support for Morales in the 2005 general election, we found some interesting results (see Table 6). In a model looking at all three departments, literacy still mattered (and in the same way). Not surprisingly, support for Morales significantly affected support for the MAS prefect candidate, but each one percent increase in support for Morales only corresponded to a percent increase in support for the MAS candidate. In this model ruralness also mattered, though it was only slightly significant and had a weak effect. The other interesting result is that in the Chuquisaca-only model, support for Morales produced an almost one-to-one increase in support for the MAS prefect candidate, but indigeneity and literacy lost any significant effect. Peculiarly, support for Morales had no effect on support for MAS prefect candidates in Santa Cruz and Tarija. The various 2005 models suggest an interesting, counter-intuitive view of voters in Santa Cruz and Tarija. While indigeneity (ethnic self-identification as indigenous) matters in Chuquisaca, it does not matter in Santa Cruz or Tarija. Despite having a 11

14 number of municipalities that are highly indigenous, and a broad range of difference along this dimension between municipalities, ethnic identity cannot explain voting differences across municipalities in Santa Cruz and Tarija. Of our socioeconomic factors, only literacy consistently predicts voting difference, by decreasing support for Morales and MAS in our two Media Luna departments. The 2006 Constituent Assembly Election and Autonomy Referendum Turning to the 2006 constituent assembly election and autonomy referendum, we see similar patterns. As with votes for MAS prefect candidates in 2005, ruralness again matters in models with all three departments, but drops out in separated models (see Table 7). Again, literacy has a significant and dramatic impact on votes for MAS in Santa Cruz and Tarija, with every one percent increase in municipal literacy rates translating into an almost three percent decrease in support for MAS. Indigeneity again matters in Chuquisaca, but not in three-department models. Table 7. Regression estimates of support for MAS in the 2006 constituent assembly election at the municipal level Santa Cruz, Tarija, and Chuquisaca Santa Cruz and Tarija Chuquisaca Rural ** Poor Literacy ** ** Indigenous ** Constant ** ** R-squared N observations Looking at support for autonomy across the three departments yields similar results (see Table 8). Ruralness and literacy matter in our three-department model, but not indigeneity or poverty. In our Santa Cruz and Tarija model, literacay again matters, and to the same degree, but ruralness has a significant negative impact on support for autonomy (though with a very small coefficient). Indigeneity again matters in Chuquisaca, decreasing support for autonomy in roughly similar proportion to its effect on support for MAS. 12

15 Table 8. Regression estimates of support for autonomy in the 2006 referendum at the municipal level Santa Cruz, Tarija, and Chuquisaca Santa Cruz and Tarija Chuquisaca Rural ** * Poor Literacy ** ** Indigenous ** Constant * * R-squared N observations Table 9. Regression estimates of support for autonomy in the 2006 referendum at the municipal level in Santa Cruz and Tarija Model A Model B Model C Rural Poor Literacy ** Indigenous Support for Evo on presidential ballot Support for opposition on ** 2005 prefect ballot Support for MAS on ** constituent assembly vote Constant * ** R-squared N observations We looked more closely at the autonomy votes in Santa Cruz and Tarija, seeking to control for political-electoral variables (see Table 9). Here, we found some interesting results. We found no evidence that support for Morales in the 2005 election could explain votes for or against autonomy. However, we found that support for the opposition prefect did have a significant impact, wiping out any literacy effect. We also found that support for MAS in the 2006 constituent assembly election also could explain autonomy votes, with a nearly one-to-one inverse correlation. If voters in Santa Cruz and Tarija had viewed their autonomy referendum votes as a reflection of their views on Morals, we would have expected a strong relationship between the two. Instead, our regression analysis suggests that support for Morales in 2005 had little impact on how voters cast 13

16 their ballots in the autonomy referendum only six months later. Coupled with our earlier findings that support for Morales in 2005 had no effect on how voters in Santa Cruz and Tarija cast their ballots for prefect (see Table 6), the evidence suggests that Morales supporters were about as likely to vote for opposition prefects and/or autonomy as Morales opponents. The 2008 Regional Autonomy Referendums in Santa Cruz and Tarija When we jump forward two years to the "wildcat" 2008 autonomy referendums held in Santa Cruz and Tarija, the results are astonishingly similar (see Table 10). In these referendums, none of the socioeconomic variables had any significant effect with the exception of poverty, which had a greater than one-to-one decreasing affect on support for autonomy, but only in Tarija. Municipalities in Tarija that had higher indexes of poverty were less likely to vote for autonomy, which as in keeping with our hypothesized projections. Table 10. Regression estimates of support for autonomy in the 2008 referendums at the municipal level in Santa Cruz and Tarija Santa Cruz and Tarija Santa Cruz Tarija Rural Poor * Literacy Indigenous Constant R-squared N observations Looking at models including political-economic variables (see Table 11), however, were consistent our previous findings. While support for autonomy in the 2006 referendum was significantly correlated with support for autonomy in But the coefficient is small, meaning that municipalities that supported autonomy in 2006 were (when controlling for socioeconomic and demographic differences) only slightly more likely to support autonomy in Interestingly, support for Morales in 2005 or the incumbent prefect in the 2005 prefect election had no significant effect on support for autonomy in Whatever drives support for autonomy in Santa Cruz and Tarija, our models offers little support for the hypothesis that indigeneity, ruralness, poverty, or literacy rates matter except for in Tarija, where only poverty matters. 14

17 Table 11. Regression estimates of support for autonomy in the 2008 referendums at the municipal level in Santa Cruz and Tarija Model A Model B Model C Rural Poor Literacy Indigenous Support for autonomy in ** 2006 referendum Support for incumbent prefect in 2005 Support for Evo in Constant R-squared N observations Because the 2008 autonomy referendums were unofficial elections and faced calls for boycotts from MAS supporters we decided to test whether voter turnout was affected by any socioeconomic, demographic, or political-electoral factors (see Table 12). Our first model, which includes no political-electoral variables, suggests that literacy had a strong effect, driving up voter turnout, but that other socioeconomic factors and indigeneity had no significant effect on voter turnout. When looking at support for the incumbent prefect, we did find a slightly greater than one-to-one correlation with voter turnout, which also wiped out any effect from literacy. When looking at support for Morales in 2005, however, we found no significant effect. This suggests that the 2008 autonomy referendum may be seen as a reflection of support for the prefect (Costas in Santa Cruz, Cossío in Tarija), but not on Morales. Table 12. Regression estimates of voter turnout in the 2008 referendums at the municipal level in Santa Cruz and Tarija Model A Model B Model C Rural Poor Literacy * * Indigenous Support for incumbent ** prefect in 2005 Support for Evo in Constant R-squared N observations

18 The 2008 Chuquisaca Prefect Election The special by election to fill the vacant Chuquisaca prefecture also offered us an interesting opportunity to look more closely at Chuquisaca, which had only since 2006 developed a regional political movement though focused on the specific "capitalía" demand of the city of Sucre. As we discuss earlier in this paper, the election was unique because it pitted an indigenous female opposition candidate against a mestizo male MAS candidate. Nevertheless, we expected that rural, poor, indigenous Chuquisaca voters would remain loyal to MAS (after all, Cuellar lost in every rural municipality). Our statistical analysis confirms most of our predictions (see Table 13). Indigenous and poor municipalities were less likely to vote for Cuellar, though poor municipalities did so by much larger margins. Interestingly, however, the effect of indigeneity drops out in models that include either support for autonomy in 2006 or support for Morales in 2005 (though poverty continues to matter). Of course, earlier models showed that indigeneity clearly matters in Chuquisaca. But it is interesting to note that when including support for Morales or MAS, the significance of indigeneity disappears (and in highly robust models). Table 13. Regression estimates of support for Savina Cuellar in the 2008 prefect election at the municipal level in Chuquisaca Model A Model B Model C Rural Poor * ** ** Literacy Indigenous ** Support for autonomy in ** 2006 referendum Support for Evo in ** Constant ** ** R-squared N observations The 2008 President and Prefect Recall Referendums Turning to the recall referendums, we tested a number of different models. When looking at support for Morales in the 2008 presidential recall referendum (see Table 14), we found that in models including all three departments, ruralness and literacy mattered, and in the expected directions: Rural municipalities were more likely to support Morales (though by very minor margins), while more literate municipalities were more likely to oppose him (by nearly one-to-one ratios). Interestingly, support for Morales in 2005 had no significant effect (when controlling for the effects of socioeconomic and demographic variables) on support for Morales three years later later. When looking only at Santa Cruz and Tarija, we found that ruralness dropped out, but support for autonomy was significant. But in this model literacy again mattered but now with an almost two-to-one 16

19 effect (a one percent increase in literacy rates produced an estimated percent decrease in support for Morales at the municipal level). Table 14. Regression estimates of support for Evo Morales in the 2008 recall referendum at the municipal level Santa Cruz, Tarija, and Chuquisaca (Model A) Santa Cruz, Tarija, and Chuquisaca (Model B) Santa Cruz and Tarija Rural ** ** Poor Literacy ** ** * Indigenous Support for autonomy in ** 2008 referendum Support for Evo in Constant ** ** ** R-squared N observations Table 15. Regression estimates of support for incumbent prefects in the 2008 recall referendum at the municipal level in Santa Cruz and Tarija Model A Model B Model C Rural Poor Literacy ** * Indigenous Support for autonomy in ** 2008 referendum Support for Evo in ** recall referendum Constant * * ** R-squared N observations When turning to support for the incumbent prefects in the 2008 recall referendums in Santa Cruz and Tarija (no such referendum took place in Chuquisaca), we found that literacy mattered (and with a very high coefficient), but so too did both our political-electoral variables (see Table 15). Our finding that municipalities that supported autonomy were more likely to support the incumbent prefect in Santa Cruz and Tarija (see Table 12 above) were again confirmed. This time, however, literacy continued to 17

20 matter and the magnitude of the coefficient for the autonomy-prefect relationship was sharply reduced. Also, these regional referendums were the first in which support for Morales was significant: municipalities that were likely to support Morales were less likely to support the incumbent prefect (and by a one-to-one ratio) and differences in literacy rates no longer mattered. This final model is noteworthy, because it is one of the most robust (an R-square value of ), suggesting that, at the municipal levels, support for Morales and the incumbent prefect were mirror images of each other. The 2009 Constitutional Referendum Finally, we turned our attention to the January 2009 constitutional referendum, which was viewed as a referendum on Morales's presidency, or the strength of regional autonomy movements, or both. We tested both of these relations (see Table 16). In models excluding political-electoral variables, we found that ruralness and literacy mattered, consistent with some of our previous findings. More literate municipalities were less likely to support the proposed constitution and rural municipalities were more likely to support it, though the effect of ruralness was quite small. When we included support for Morales in the 2008 recall referendum, we confirmed the expectation that Morales supporters were likely to support the proposed constitution (and with a nearly one-to-one ratio). Including support for Morales, however, wiped out the ruralness effect and increased the literacy effect it also introduced a significant (though very slight) indigeneity effect. Looking at support for autonomy in the 2008 referendum (which took place about six months earlier) produced similar results: indigeneity again mattered (and to a comparably larger degree), as did literacy (and by a dramatically degree). Still, this model was not as robust as the model that looked at support for Morales in the recall referendum; that model could explain nearly 98 percent of all variation across the 89 observed municipalities. Table 16. Regression estimates of support for constitution in the 2009 referendum at the municipal level Model A Model B Model C Rural ** Poor Literacy ** ** ** Indigenous ** * Support for autonomy in ** 2008 referendum Support for Evo in ** recall referendum Constant ** ** R-squared N observations

21 Limitations It is possible that our analysis missed some important socioeconomic and/or demographic variables. For example, it is possible that ethnic identity in Santa Cruz and Tarija is more complex than in Chuquisaca. After all, Chuquisaca not only has a larger total indigenous population (65.5 percent), but most of this is made up of a single group (those who self-identify as Quechua comprise 61 percent of the population). In contrast, only 37.4 percent of Santa Cruz residents who self-identified as belonging to an indigenous community in the 2001 census, and these are divided into five major ethnic groups (though nearly half of the total self-identify as Quechua), including 2.2 percent other. Tarija has only 19.7 percent indigenous population, more than half of whom identify as Quechua, but includes an additional four other major groups (and 1.5 percent other ). It is also possible that models that took into account differences between ethnic indigenous groups might find that certain groups (e.g. Quechua) might more closely identify with Morales than other groups. We are certainly sensitive to the diversity within Bolivia s indigenous community. Unfortunately, we have not yet had an opportunity to disaggregate our data on ethnic differences at the municipal levels. We hope to do so in the near future. Conclusion Overall, our conclusions are counter-intuitive. Much of the statistical analysis discussed above suggests that across the three departments included in this study, socioeconomic and demographic factors are not powerful predictors of differences in support for Evo Morales and MAS. Despite significant numbers of poor, rural, and indigenous people living in the three departments and the broad variation in their distribution across municipalities support for Morales and MAS in the early years under observation is difficult to predict solely on socioeconomic or demographic factors. Two significant exceptions are Chuquisaca, where indigeneity mattered, and in Santa Cruz and Tarija, where literacy rates mattered a great deal. Socioeconomic factors begin to matter across the three cases only in the later years under observation, and primarily in models that include previous support for Morales. In fact, indigeneity only matters across all three departments in the most recent 2009 constitutional referendum, though only in models that include previous support for Morales and/or autonomy. Despite the domestic and international recognition of Evo Morales as the first indigenous president of Bolivia, and the recognition that he campaigned on a platform that sought to appeal to poor, rural, and other underprivileged sectors of Bolivia's population, we found little evidence to suggest that these factors played any significant role in mobilizing voters in favor of (or in opposition to) Morales or MAS in 2005 and 2006 in Santa Cruz and Tarija. What did seem to matter in these two departments was literacy and this effect increased in magnitude by 2008 and Indigeneity did matter in Chuquisaca, however, though this effect generally declined in magnitude by 2008 and Whatever drove voters to back or oppose Morales in Santa Cruz and Tarija across municipalities (other than literacy rates) is not captured by our socioeconomic or demographic indicators. A previous study by Centellas (2005) has suggested that voting differences across Bolivia s provinces could be explained by a regional ( Media Luna ) dummy variable. We would like to further test our analysis by expanding our cases to 19